Role of artificial intelligence applied to ultrasound in gynecology oncology: A systematic review.

The aim of this paper was to explore the role of artificial intelligence (AI) applied to ultrasound imaging in gynecology oncology. Web of Science, PubMed, and Scopus databases were searched. All studies were imported to RAYYAN QCRI software. The overall quality of the included studies was assessed using QUADAS-AI tool. Fifty studies were included, of these 37/50 (74.0%) on ovarian masses or ovarian cancer, 5/50 (10.0%) on endometrial cancer, 5/50 (10.0%) on cervical cancer, and 3/50 (6.0%) on other malignancies. Most studies were at high risk of bias for subject selection (i.e., sample size, source, or scanner model were not specified; data were not derived from open-source datasets; imaging preprocessing was not performed) and index test (AI models was not externally validated) and at low risk of bias for reference standard (i.e., the reference standard correctly classified the target condition) and workflow (i.e., the time between index test and reference standard was reasonable). Most studies presented machine learning models (33/50, 66.0%) for the diagnosis and histopathological correlation of ovarian masses, while others focused on automatic segmentation, reproducibility of radiomics features, improvement of image quality, prediction of therapy resistance, progression-free survival, and genetic mutation. The current evidence supports the role of AI as a complementary clinical and research tool in diagnosis, patient stratification, and prediction of histopathological correlation in gynecological malignancies. For example, the high performance of AI models to discriminate between benign and malignant ovarian masses or to predict their specific histology can improve the diagnostic accuracy of imaging methods.

A clinical ultrasound algorithm to identify uterine sarcoma and smooth muscle tumors of uncertain malignant potential (STUMPs) in patients with myometrial lesions: the MYLUNAR Study.

BACKGROUND: Differential diagnosis between benign uterine smooth muscle tumors and malignant counterpart is challenging. We evaluated the accuracy of a clinical and ultrasound based algorithm in predicting mesenchimal uterine malignancies (MUMs), including smooth muscle tumors of uncertain malignant potential (STUMPs). METHODS: We report the twelve-months follow-up of an observational, prospective, single-centre study that included women with at least one myometrial lesion ≥3 cm on ultrasound examination. These patients were classified according to a three-class diagnostic algorithm, using symptoms and ultrasound features. "White" patients underwent annual telephone follow-up for 2 years, "Green" patients underwent a clinical and ultrasound follow-up at 6, 12 and 24 months and "Orange" patients underwent surgery. We further developed a risk class system to stratify the malignancy risk. FINDINGS: 2,268 women were included andtarget lesion was classified as benign in 2,158 (95.1%), as other malignancies in 58 (2.6%) an as mesenchymal uterine malignancies in 52 (2.3%) patients. At multivariable analysis, age (OR 1.05 (95% CI 1.03-1.07), tumor diameter >8 cm (OR 5.92 (95% CI 2.87-12.24), irregular margins (OR 2.34 (95% CI 1.09-4.98), color score=4 (OR 2.73 (95% CI 1.28-5.82), were identified as independent risk factors for malignancies, whereas acoustic shadow resulted in an independent protective factor (OR 0.39 (95% CI 0.19-0.82). The model, which included age as a continuous variable and lesion diameter as a dichotomized variable (cut-off 81 mm), provided the best AUC (0.87 (95% CI 0.82-0.91)). A risk class system was developed, and patients were classified as low-risk (predictive model value <0.39%: 0/606 malignancies, risk 0%), intermediate risk (predictive model value 0.40%-2.2%: 9/1,093 malignancies, risk 0.8%), high risk (predictive model value ≥2.3%: 43/566 malignancies, risk 7.6%). CONCLUSION: The preoperative three-class diagnostic algorithm and risk class system can stratify women according to risk of malignancy. Our findings, if confirmed in a multicentre study, will permit differentiation between benign and MUMs allowing a personalized clinical approach. FUNDING: Nothing to declare.

Prediction of non-resectability using the updated Predictive Index Value model assessed by imaging and surgery in tubo-ovarian cancer: a prospective multicenter ISAAC study.

BACKGROUND: A laparoscopy-based scoring system was developed by Fagotti et al (Fagotti or Predictive Index Value (PIV)score) based on the intraoperative presence or absence of carcinomatosis on predefined sites. Later, the authors updated the PIV score calculated only in the absence of one or both absolute criteria of non-resectability (mesenteric retraction and miliary carcinomatosis of the small bowel) (updated PIV model). OBJECTIVE: The aim was to demonstrate the non-inferiority of ultrasound to other imaging methods (contrast enhanced computed tomography (CT) and whole-body diffusion-weighted (WB DWI)/MRI) in predicting non-resectable tumor (defined as residual disease>1 cm) using the updated PIV model in patients with tubo-ovarian cancer. The agreement between imaging and intraoperative findings as a reference was also calculated. STUDY DESIGN: This was a European prospective multicenter observational study. We included patients with suspected tubo-ovarian carcinoma who underwent preoperative staging and prediction of non-resectability at ultrasound, CT, WB-DWI/MRI and surgical exploration. The predictors of non-resectability were suspicious mesenteric retraction and/or miliary carcinomatosis of the small bowel or if absent, a PIV>8 (updated PIV model). The PIV score ranges from 0 to 12 according to the presence of disease in six predefined intra-abdominal sites (great omentum, liver surface, lesser omentum/stomach/spleen, parietal peritoneum, diaphragms, bowel serosa/mesentery). The reference standard was surgical outcome, in terms of residual disease>1 cm, assessed by laparoscopy and/or laparotomy. The area under the receiver operating characteristic curve (AUC) to assess the performance of the methods in predicting non-resectability was reported. Concordance between index tests at detection of disease at six predefined sites and intraoperative exploration as reference standard was also calculated using Cohen's kappa. RESULTS: The study was between 2018 and 2022 in five European gynecological oncology centers. Data from 242 patients having both mandatory index tests (ultrasound and CT) were analyzed. 145/242 (59.9%) patients had no macroscopic residual tumor after surgery (R0) (5/145 laparoscopy and 140/145 laparotomy) and 17/242 (7.0%) had residual tumor ≤1cm (R1) (laparotomy). In 80/242 patients (33.1%), the residual tumor was >1 cm (R2), 30 of them underwent laparotomy and maximum surgery was carried out and 50/80 underwent laparoscopy and cytoreduction was not feasible in all of them. After excluding 18/242 (7.4%) patients operated on but not eligible for extensive surgery, the predictive performance of three imaging methods was analyzed in 167 women. The AUCs of all methods in discriminating between resectable and non-resectable tumor was 0.80 for ultrasound, 0.76 for CT, 0.71 for WB-DWI/MRI and 0.90 for surgical exploration. Ultrasound had the highest agreement (Cohen's kappa ranging from 0.59 to 0.79) compared to CT and WB-DWI/MRI to assess all parameters included in the updated PIV model. CONCLUSIONS: Ultrasound showed non-inferiority to CT and to WB-DWI/MRI in discriminating between resectable and non-resectable tumor using the updated PIV model. Ultrasound had the best agreement between imaging and intraoperative findings in the assessment of parameters included in the updated PIV model. Ultrasound is an acceptable method to assess abdominal disease and predict non-resectability in patients with tubo-ovarian cancer in the hands of specially trained ultrasound examiners.

Patient satisfaction with ultrasound, whole-body CT and whole-body diffusion-weighted MRI for pre-operative ovarian cancer staging: a multicenter prospective cross-sectional survey.

BACKGROUND: In addition to the diagnostic accuracy of imaging methods, patient-reported satisfaction with imaging methods is important. OBJECTIVE: To report a secondary outcome of the prospective international multicenter Imaging Study in Advanced ovArian Cancer (ISAAC Study), detailing patients' experience with abdomino-pelvic ultrasound, whole-body contrast-enhanced computed tomography (CT), and whole-body diffusion-weighted magnetic resonance imaging (WB-DWI/MRI) for pre-operative ovarian cancer work-up. METHODS: In total, 144 patients with suspected ovarian cancer at four institutions in two countries (Italy, Czech Republic) underwent ultrasound, CT, and WB-DWI/MRI for pre-operative work-up between January 2020 and November 2022. After having undergone all three examinations, the patients filled in a questionnaire evaluating their overall experience and experience in five domains: preparation before the examination, duration of examination, noise during the procedure, radiation load of CT, and surrounding space. Pain perception, examination-related patient-perceived unexpected, unpleasant, or dangerous events ('adverse events'), and preferred method were also noted. RESULTS: Ultrasound was the preferred method by 49% (70/144) of responders, followed by CT (38%, 55/144), and WB-DWI/MRI (13%, 19/144) (p<0.001). The poorest experience in all domains was reported for WB-DWI/MRI, which was also associated with the largest number of patients who reported adverse events (eg, dyspnea). Patients reported higher levels of pain during the ultrasound examination than during CT and WB-DWI/MRI (p<0.001): 78% (112/144) reported no pain or mild pain, 19% (27/144) moderate pain, and 3% (5/144) reported severe pain (pain score >7 of 10) during the ultrasound examination. We did not identify any factors related to patients' preferred method. CONCLUSION: Ultrasound was the imaging method preferred by most patients despite being associated with more pain during the examination in comparison with CT and WB-DWI/MRI. TRIAL REGISTRATION NUMBER: NCT03808792.

Innovative Ultrasound Criteria for the Diagnosis of Adenomyosis and Correlation with Symptoms: A Retrospective Re-Evaluation.

The 2022 Delphi revision of the MUSA (Morphological Uterus Sonographic Assessment) criteria for the ultrasound diagnosis of adenomyosis divides the ultrasound signs for diagnosis into direct and indirect ones, considering the presence of at least one direct sign as a mandatory criterion. This study aimed to reclassify the patients referred to the Pelvic Pain specialist outpatient clinic of the Gynecological Clinic of Udine according to the new criteria, evaluating the number of overdiagnoses and the possible correlation between the direct and indirect signs and the patients' symptoms. 62 patients affected by adenomyosis were retrospectively recruited. The patients were then re-evaluated by ultrasound and clinically. At least one direct sign of adenomyosis was found in 52 patients, while 16% of the population examined did not present any. There was no statistically significant difference between patients presenting direct signs and those presenting none for the symptoms considered. According to the new criteria, 16% of the patients examined were not affected by adenomyosis; applying the new consensus to symptomatic patients could increase false negatives. In a population of symptomatic patients, the diagnosis of adenomyosis is still highly probable even without direct ultrasound signs, given the clinical symptoms and having ruled out other causes of such symptoms.

Comparison of clinical and ultrasound examinations in assessing the parametria in patients with deep infiltrating endometriosis: a multicentre prospective study.

RESEARCH QUESTION: How do clinical rectovaginal examination and transvaginal ultrasound examination perform in the diagnosis of parametrial infiltration in patients with endometriosis? DESIGN: This was a multicentre prospective observational study. Patients with suspected deep endometriosis at clinical examination and/or at ultrasound evaluation and scheduled for surgery were included. Following multicentre multidisciplinary meetings, consensus was obtained on terms and methodology to define the parametrium at pelvic anatomy, ultrasound and surgery. Sensitivity, specificity, accuracy, and positive and negative likelihood ratios were calculated for clinical and ultrasound examinations with respect to surgery. RESULTS: In total, 195 women were selected for the present study and 164 were included in the analysis. Ultrasound examination had good to high specificity (>80%) for all parameters, except the left lateral parametrium (78.8%). The sensitivity of ultrasound examination was good to high for fixity of the right and left ovaries, uterosacral ligaments, retrocervix and rectovaginal space; and low for the anterior and lateral parametria, vagina, bladder and bowel. Clinical examination had good to high specificity for fixity of the left ovary, anterior parametrium, right uterosacral ligament, retrocervix and vagina; and low specificity for fixity of the right ovary, lateral parametrium, left uterosacral ligament and rectovaginal space. The sensitivity of clinical examination was good for the uterosacral ligaments and rectovaginal space, and low for the remaining parameters. CONCLUSION: Ultrasound examination provided good specificity for all the parameters, but sensitivity was low for the anterior and lateral parametria. Clinical examination provided good specificity for the anterior and posterior parametria, but sensitivity was low for the anterior and lateral parametria. Further prospective studies are needed to validate this methodology and confirm the results.

Comparison of the ADNEX and ROMA risk prediction models for the diagnosis of ovarian cancer: a multicentre external validation in patients who underwent surgery.

BACKGROUND: Several diagnostic prediction models to help clinicians discriminate between benign and malignant adnexal masses are available. This study is a head-to-head comparison of the performance of the Assessment of Different NEoplasias in the adneXa (ADNEX) model with that of the Risk of Ovarian Malignancy Algorithm (ROMA). METHODS: This is a retrospective study based on prospectively included consecutive women with an adnexal tumour scheduled for surgery at five oncology centres and one non-oncology centre in four countries between 2015 and 2019. The reference standard was histology. Model performance for ADNEX and ROMA was evaluated regarding discrimination, calibration, and clinical utility. RESULTS: The primary analysis included 894 patients, of whom 434 (49%) had a malignant tumour. The area under the receiver operating characteristic curve (AUC) was 0.92 (95% CI 0.88-0.95) for ADNEX with CA125, 0.90 (0.84-0.94) for ADNEX without CA125, and 0.85 (0.80-0.89) for ROMA. ROMA, and to a lesser extent ADNEX, underestimated the risk of malignancy. Clinical utility was highest for ADNEX. ROMA had no clinical utility at decision thresholds <27%. CONCLUSIONS: ADNEX had better ability to discriminate between benign and malignant adnexal tumours and higher clinical utility than ROMA. CLINICAL TRIAL REGISTRATION: clinicaltrials.gov NCT01698632 and NCT02847832.